From 55ab3a1ff843e3f0e24d2da44e71bffa5d853010 Mon Sep 17 00:00:00 2001 From: Anton Blanchard Date: Tue, 6 Apr 2010 14:34:58 -0700 Subject: raw: fsync method is now required Commit 148f948ba877f4d3cdef036b1ff6d9f68986706a (vfs: Introduce new helpers for syncing after writing to O_SYNC file or IS_SYNC inode) broke the raw driver. We now call through generic_file_aio_write -> generic_write_sync -> vfs_fsync_range. vfs_fsync_range has: if (!fop || !fop->fsync) { ret = -EINVAL; goto out; } But drivers/char/raw.c doesn't set an fsync method. We have two options: fix it or remove the raw driver completely. I'm happy to do either, the fact this has been broken for so long suggests it is rarely used. The patch below adds an fsync method to the raw driver. My knowledge of the block layer is pretty sketchy so this could do with a once over. If we instead decide to remove the raw driver, this patch might still be useful as a backport to 2.6.33 and 2.6.32. Signed-off-by: Anton Blanchard Reviewed-by: Jan Kara Cc: Christoph Hellwig Cc: Alexander Viro Cc: Jens Axboe Reviewed-by: Jeff Moyer Tested-by: Jeff Moyer Cc: Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/block_dev.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'fs/block_dev.c') diff --git a/fs/block_dev.c b/fs/block_dev.c index d11d0289f3d2..8db62b2b6df8 100644 --- a/fs/block_dev.c +++ b/fs/block_dev.c @@ -404,7 +404,7 @@ static loff_t block_llseek(struct file *file, loff_t offset, int origin) * NULL first argument is nfsd_sync_dir() and that's not a directory. */ -static int block_fsync(struct file *filp, struct dentry *dentry, int datasync) +int block_fsync(struct file *filp, struct dentry *dentry, int datasync) { struct block_device *bdev = I_BDEV(filp->f_mapping->host); int error; @@ -418,6 +418,7 @@ static int block_fsync(struct file *filp, struct dentry *dentry, int datasync) error = 0; return error; } +EXPORT_SYMBOL(block_fsync); /* * pseudo-fs -- cgit v1.2.1 From b1dd3b2843b3b73b7fc2ee47d96310cd1c051371 Mon Sep 17 00:00:00 2001 From: Andrew Morton Date: Tue, 6 Apr 2010 14:35:00 -0700 Subject: vfs: rename block_fsync() to blkdev_fsync() Requested by hch, for consistency now it is exported. Cc: Alexander Viro Cc: Anton Blanchard Cc: Christoph Hellwig Cc: Jan Kara Cc: Jeff Moyer Cc: Jens Axboe Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/block_dev.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'fs/block_dev.c') diff --git a/fs/block_dev.c b/fs/block_dev.c index 8db62b2b6df8..2a6d0193f139 100644 --- a/fs/block_dev.c +++ b/fs/block_dev.c @@ -404,7 +404,7 @@ static loff_t block_llseek(struct file *file, loff_t offset, int origin) * NULL first argument is nfsd_sync_dir() and that's not a directory. */ -int block_fsync(struct file *filp, struct dentry *dentry, int datasync) +int blkdev_fsync(struct file *filp, struct dentry *dentry, int datasync) { struct block_device *bdev = I_BDEV(filp->f_mapping->host); int error; @@ -418,7 +418,7 @@ int block_fsync(struct file *filp, struct dentry *dentry, int datasync) error = 0; return error; } -EXPORT_SYMBOL(block_fsync); +EXPORT_SYMBOL(blkdev_fsync); /* * pseudo-fs @@ -1482,7 +1482,7 @@ const struct file_operations def_blk_fops = { .aio_read = generic_file_aio_read, .aio_write = blkdev_aio_write, .mmap = generic_file_mmap, - .fsync = block_fsync, + .fsync = blkdev_fsync, .unlocked_ioctl = block_ioctl, #ifdef CONFIG_COMPAT .compat_ioctl = compat_blkdev_ioctl, -- cgit v1.2.1 From b8af67e2681c693a21f3933e3bdfce4cf66596d3 Mon Sep 17 00:00:00 2001 From: Anton Blanchard Date: Fri, 23 Apr 2010 13:18:06 -0400 Subject: fs/block_dev.c: fix performance regression in O_DIRECT|O_SYNC writes to block devices We are seeing a large regression in database performance on recent kernels. The database opens a block device with O_DIRECT|O_SYNC and a number of threads write to different regions of the file at the same time. A simple test case is below. I haven't defined DEVICE since getting it wrong will destroy your data :) On an 3 disk LVM with a 64k chunk size we see about 17MB/sec and only a few threads in IO wait: procs -----io---- -system-- -----cpu------ r b bi bo in cs us sy id wa st 0 3 0 16170 656 2259 0 0 86 14 0 0 2 0 16704 695 2408 0 0 92 8 0 0 2 0 17308 744 2653 0 0 86 14 0 0 2 0 17933 759 2777 0 0 89 10 0 Most threads are blocking in vfs_fsync_range, which has: mutex_lock(&mapping->host->i_mutex); err = fop->fsync(file, dentry, datasync); if (!ret) ret = err; mutex_unlock(&mapping->host->i_mutex); commit 148f948ba877f4d3cdef036b1ff6d9f68986706a (vfs: Introduce new helpers for syncing after writing to O_SYNC file or IS_SYNC inode) offers some explanation of what is going on: Use these new helpers for syncing from generic VFS functions. This makes O_SYNC writes to block devices acquire i_mutex for syncing. If we really care about this, we can make block_fsync() drop the i_mutex and reacquire it before it returns. Thanks Jan for such a good commit message! As well as dropping i_mutex, Christoph suggests we should remove the call to sync_blockdev(): > sync_blockdev is an overcomplicated alias for filemap_write_and_wait on > the block device inode, which is exactly what we did just before calling > into ->fsync The patch below incorporates both suggestions. With it the testcase improves from 17MB/s to 68M/sec: procs -----io---- -system-- -----cpu------ r b bi bo in cs us sy id wa st 0 7 0 65536 1000 3878 0 0 70 30 0 0 34 0 69632 1016 3921 0 1 46 53 0 0 57 0 69632 1000 3921 0 0 55 45 0 0 53 0 69640 754 4111 0 0 81 19 0 Testcase: #define _GNU_SOURCE #include #include #include #include #include #include #include #include #define NR_THREADS 64 #define BUFSIZE (64 * 1024) #define DEVICE "/dev/mapper/XXXXXX" #define ALIGN(VAL, SIZE) (((VAL)+(SIZE)-1) & ~((SIZE)-1)) static int fd; static void *doit(void *arg) { unsigned long offset = (long)arg; char *b, *buf; b = malloc(BUFSIZE + 1024); buf = (char *)ALIGN((unsigned long)b, 1024); memset(buf, 0, BUFSIZE); while (1) pwrite(fd, buf, BUFSIZE, offset); } int main(int argc, char *argv[]) { int flags = O_RDWR|O_DIRECT; int i; unsigned long offset = 0; if (argc > 1 && !strcmp(argv[1], "O_SYNC")) flags |= O_SYNC; fd = open(DEVICE, flags); if (fd == -1) { perror("open"); exit(1); } for (i = 0; i < NR_THREADS-1; i++) { pthread_t tid; pthread_create(&tid, NULL, doit, (void *)offset); offset += BUFSIZE; } doit((void *)offset); return 0; } Signed-off-by: Anton Blanchard Acked-by: Jan Kara Cc: Christoph Hellwig Cc: Alexander Viro Cc: Jens Axboe Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/block_dev.c | 17 ++++++++++++----- 1 file changed, 12 insertions(+), 5 deletions(-) (limited to 'fs/block_dev.c') diff --git a/fs/block_dev.c b/fs/block_dev.c index 2a6d0193f139..6dcee88c2e5d 100644 --- a/fs/block_dev.c +++ b/fs/block_dev.c @@ -406,16 +406,23 @@ static loff_t block_llseek(struct file *file, loff_t offset, int origin) int blkdev_fsync(struct file *filp, struct dentry *dentry, int datasync) { - struct block_device *bdev = I_BDEV(filp->f_mapping->host); + struct inode *bd_inode = filp->f_mapping->host; + struct block_device *bdev = I_BDEV(bd_inode); int error; - error = sync_blockdev(bdev); - if (error) - return error; - + /* + * There is no need to serialise calls to blkdev_issue_flush with + * i_mutex and doing so causes performance issues with concurrent + * O_SYNC writers to a block device. + */ + mutex_unlock(&bd_inode->i_mutex); + error = blkdev_issue_flush(bdev, NULL); if (error == -EOPNOTSUPP) error = 0; + + mutex_lock(&bd_inode->i_mutex); + return error; } EXPORT_SYMBOL(blkdev_fsync); -- cgit v1.2.1